JARVIS. ONLINE · OPUS 4.8
THE FLEET · NEURAL LINK
AGENTS
8 specialists run your operation. Jarvis chairs the board; each agent owns one job and reports back to the core.
JARVIS
Chief of Staff · orchestrator · the core
The orchestrator. Chairs the morning board, reviews yesterday, sets today's top three priorities by revenue, and assigns every other agent. Nothing ships without passing through it.
ANALYST
Response Analyst
Tracks how long every customer has been waiting and ranks the day by revenue impact — so the team always works the most valuable thread first.
CLOSER
The Closer
Drafts replies in your voice on the highest-value waiting threads — ready to review and send. Never sends on its own.
CHASER
The Chaser
Hunts stale and dropped threads — deals that went quiet — and drafts the follow-up nudge to bring them back to life.
COACH
Sales Manager
Coaches you and the team off your real sent replies — what's landing, where to tighten up, and who needs help this week.
BUILDER
Developer
Keeps the dashboard and the draft pipeline healthy, and auto-fixes Tier-1 bugs before they ever reach you.
OPTIMIZER
The Optimizer
Audits the systems and software that already work — hunting for ways to make them faster, cheaper, and more accurate — and proposes the highest-leverage wins for approval.
JANITOR
Janitor
Safe housekeeping and system hygiene — clears clutter, archives the noise, and keeps everything tidy and fast.
APPROVAL QUEUE
Decisions the fleet has staged for you. Approve to let them run, or skip.
▸ PENDING AUTHORIZATION41
Janitor
SLA audit: 3 breached, 237 stale drafts
Nightly inbox SLA audit over the last 7 days found: 3 received emails past their respond-by deadline, 1 tier-1 emails still open (status received), and 237 drafted emails sitting more than 12 hours past respond-by. Check the emails table (status, tier, respond_by) and the Draft Pipeline executions in n8n to see which threads need a reply sent or a draft finished.
Opportunity
Build a category-volume snapshot into daily_metrics
Revenue thread categories (271/125/108/38/20/12) are visible in the live snapshot but not persisted, so shifts in inquiry vs. logistics mix can't be tracked over time. Capturing the distribution nightly reveals demand trends and pipeline bottlenecks.
Opportunity
Reconcile the 20 Pricing threads against a rate-card lookup
Pricing (20) and ETA/shipping (12) are small but high-friction categories where drafts likely guess at numbers, risking wrong quotes. Grounding these drafts in a structured rate card eliminates the most error-prone manual replies.
Opportunity
Add a stuck-card aging timer to the 50-card approval queue
50 pending approval cards is at the upper end of healthy backlog and cards can sit for days, silently delaying revenue replies. Tracking how long each card has waited turns a static count into an actionable aging signal.
Opportunity
Route the 271 new inquiries into a dedicated fast-draft lane
New inquiries (271) are by far the largest revenue category and the top of the funnel, yet they flow through the same generic Draft Pipeline as logistics and feedback. A dedicated lane with inquiry-specific prompt + priority ensures the highest-conversion threads get answered first.
Opportunity
Add a draft-to-email coverage delta alarm in Night Brain
1086 emails but only 735 drafts means 351 emails (32%) have no draft, and the gap is growing silently across sweeps. A nightly delta alarm catches the drafting pipeline stalling before it compounds into hundreds of unanswered revenue threads.
Janitor
SLA audit: 13 breached, 164 stale drafts
Nightly inbox SLA audit over the last 7 days found: 13 received emails past their respond-by deadline, 2 tier-1 emails still open (status received), and 164 drafted emails sitting more than 12 hours past respond-by. Check the emails table (status, tier, respond_by) and the Draft Pipeline executions in n8n to see which threads need a reply sent or a draft finished.
Opportunity
Add a one-click batch-approve for low-risk Order logistics cards
Order logistics (130) and ETA/shipping (13) threads are routine status updates that still each require individual approval, inflating the 43-card backlog with low-stakes clicks. Batching the safe categories lets Alex clear volume in seconds.
Opportunity
Build a draft acceptance-rate metric from Capture Sent Feedback
639 drafts exist and feedback is being captured, but there's no single number showing what fraction of drafts get sent as-is versus heavily edited or killed. That ratio is the core signal of whether the Claude pipeline is actually saving Alex time.
Opportunity
Surface 76-unclassified count as a daily health KPI
76 emails (≈7.5% of 1007) carry no category, blocking every category-routed workflow downstream, but the only visibility is a static flag. A trend line catches whether classification is keeping up with intake or falling behind.